Did you start this idea of change after the first big pipeline failure? Was it a high-level view, or are you approaching it by therapy area?

When we had Exanta [the first-in-class direct thrombid inhibitor anticoagulant] and Iressa [the first-in-class tyrosine kinase inhibitor for cancer], which wasn't a complete failure but was turned down by FDA—it's still doing well in Asia, the company changed the organization. I was brought in in January 2005 to diagnose the problem and then do something about it. So the answer is, it precipitated a significant change at the top.

With regard to safety and toxicology, where we were in 2005 is worlds different from where we are today. But in your first few days, did you see the problem? Is it something you're addressing bit by bit, or did you come in with a total plan?

I followed that over the next six months with the whole project portfolio of all 70 or 80 molecules. We reviewed every single one in depth, challenging the teams to come forward and tell us what their issues were and what they were doing about them. And at the end of all that, my conclusion was that we had to be much more brutal about the quality of our molecules and deal with them on a data-driven basis, rather than on a perception basis, to make the hard decisions. And we worked to have a key decision point for every single project. And the second thing was obvious: We were just simply too slow.

So I then spent the next four or five months with my management team agreeing on how we were going to change. And we brought in something that I called Quality on Time. So everything has to have quality: quality molecules, quality programs, quality people, and quality decisions. That's the core.

And then, obviously, time. We set for ourselves the target of a median of eight years for doing development from first GLP talks through approval. We had been running at eleven and a half.

Where were you tracking on cost and cost-effectiveness?

We were not doing well. Our R&D/sales ratio was low because we had very few Phase III projects. And we were doing a lot of work in the highest-cost countries and not challenging our cost base enough. So at the end of the day, if you don't have quality molecules, you waste your money.

So you really hit the ground running. How is it all working?

It's really only in the first half of this year that we are starting to see the benefits. We've taken a huge chunk out of our development plan's time scales—our median time now is less than nine years. We're seeing progress in all elements. We're even beginning to see it in discovery, where time from lead optimization to first GLP talks has been reduced significantly.

On the issue of reduced time, other R&D chiefs we've spoken to say they're also putting a lot of effort into the discovery phase.

I've always said, "You can do a bad job with a really good molecule, and it'll probably still get to the marketplace. But if you start off with a bad molecule, it doesn't matter how good a job you do, you've wasted your time." So unless you concentrate on getting it right at the beginning, everything else falls over.

In my view, there are two issues critical to successful R&D. One is speed, and the other is the attrition rate. If you can improve attrition even a little bit, it transforms the economics of R&D in our industry. And attrition is, to some degree, inherent in the properties of the molecule that you took in.